Early Stage Detection of Scoliosis Using Machine Learning Algorithms

Scoliosis is the most common disease which is mainly identified from patient spine X-ray images. It is mainly diagnosed based on sideways curvature image modality. In scoliosis diagnosis detection, currently the treatment for scoliosis is based on spinal assessment manual study which has some limitations like it is very tedious, time-consuming and cost effective. Scoliosis diagnosis is a critical task in initial stages based on patient history records or on captured patient X-ray spine images. So our research work is carried out for detecting scoliosis in effective way by analyzing the quality of input X-ray images of the patient suffering from scoliosis. To overcome few limitations and for early stage predictive analysis detection of scoliosis, we develop a point-based automated method at different regions of spine which provides accurate results using various classification algorithms in this research paper. Predictive analysis is mainly analyzed using efficient and essential classification algorithms like Linear Regression (LR) and Support Vector Machine (SVM) and the performance metrics like Accuracy (A) and Elapsed Time (ET) are compared to benefit the patients suffering from scoliosis with less time, low cost and by improving the quality of input images.

I. Introduction

Scoliosis is represented as 3D (Three-dimensional) in structure of spine part of human body. One of the most common forms of scoliosis is AIS (Adolescent Idiopathic Scoliosis). This scoliosis form mainly begins from early stage and effects 5% of youngsters and also the cause of 75% – 80% of scoliosis cannot be identified with certainty which affects our quality of life functioning stages [ 1 – 3 ]. Based on the input spine image curve or shape or its structure, the scoliosis is classified as Normal shape, C-Curve shaped and S-Curve shaped [4] . The different shapes of Scoliosis are shown in below figure 1 :